Metric	Value
Test cosine similarity	0.88
LULC downstream accuracy	0.869
Trainable parameters	104.8M
Total parameters	104.8M
Inputs	S2 L1C+L2A (9ch), S1 RTC (2ch), COP-DEM (1ch)
Output	(H, W, 64) float32, L2-normalised

Usage

pip install betaearth

python

from betaearth import BetaEarth

model = BetaEarth.from_pretrained("asterisk-labs/betaearth-segformer")
embedding = model.predict(
    s2_l2a=s2_l2a,   # (9, H, W) uint16
    s1=s1,            # (2, H, W) float32
    dem=dem,          # (1, H, W) float32
    doy=182,
)
# embedding: (H, W, 64) numpy array

All BetaEarth models

Model	Cos Sim	Params	Best for
betaearth-segformer-film	0.886	0.3M	Best quality
betaearth-segformer-film-hilr	0.886	0.3M	Alt frozen
betaearth-segformer	0.880	104.8M	No timestamp
betaearth-segformer-film-scratch	0.883	104.8M	End-to-end
betaearth-rgb-only	0.836	26.3M	Minimal data

Citation

bibtex

@inproceedings{czerkawski2026betaearth,
  title     = {BetaEarth: Emulating Closed-Source Earth Observation Foundation Models Through Their Public Embeddings},
  author    = {Czerkawski, Mikolaj},
  booktitle = {ISPRS Congress 2026},
  year      = {2026}
}

License

CC-BY 4.0. Training data attribution: "The AlphaEarth Foundations Satellite Embedding dataset is produced by Google and Google DeepMind."